diff --git a/src/lib/core/cli.ts b/src/lib/core/cli.ts
index 3ce34b10..2a983f53 100644
--- a/src/lib/core/cli.ts
+++ b/src/lib/core/cli.ts
@@ -29,6 +29,150 @@ export interface Config {
   outputColumnName: string;
 }
 
+export type DatasetFormat = 'csv' | 'json';
+
+const DATASET_FILENAMES: Record<DatasetFormat, string> = {
+  csv: 'dataset.csv',
+  json: 'dataset.json',
+};
+
+export function detectDatasetFormat(datasetPath: string): DatasetFormat {
+  const ext = path.extname(datasetPath).toLowerCase();
+  if (ext === '.csv') {
+    return 'csv';
+  }
+  if (ext === '.json') {
+    return 'json';
+  }
+  throw new Error(`Unsupported dataset format: ${datasetPath}`);
+}
+
+function parseCsvLine(line: string): string[] {
+  const result: string[] = [];
+  let current = '';
+  let inQuotes = false;
+
+  for (let i = 0; i < line.length; i += 1) {
+    const char = line[i];
+    const next = line[i + 1];
+
+    if (inQuotes) {
+      if (char === '"' && next === '"') {
+        current += '"';
+        i += 1; // skip escaped quote
+      } else if (char === '"') {
+        inQuotes = false;
+      } else {
+        current += char;
+      }
+    } else if (char === ',') {
+      result.push(current);
+      current = '';
+    } else if (char === '"') {
+      inQuotes = true;
+    } else {
+      current += char;
+    }
+  }
+
+  result.push(current);
+  return result;
+}
+
+export function parseCsv(content: string): Record<string, string>[] {
+  const lines = content.split(/\r?\n/).filter((line) => line.trim().length > 0);
+  if (lines.length === 0) {
+    return [];
+  }
+
+  const headers = parseCsvLine(lines[0]);
+  return lines.slice(1).map((line) => {
+    const values = parseCsvLine(line);
+    const row: Record<string, string> = {};
+    headers.forEach((header, idx) => {
+      row[header] = values[idx] ?? '';
+    });
+    return row;
+  });
+}
+
+function escapeCsvValue(value: unknown): string {
+  if (value === undefined || value === null) {
+    return '';
+  }
+  const str = String(value);
+  const needsQuotes = /[",\n\r]/.test(str);
+  const escaped = str.replace(/"/g, '""');
+  return needsQuotes ? `"${escaped}"` : escaped;
+}
+
+function collectHeaders(rows: Record<string, unknown>[]): string[] {
+  const headers: string[] = [];
+  rows.forEach((row) => {
+    Object.keys(row).forEach((key) => {
+      if (!headers.includes(key)) {
+        headers.push(key);
+      }
+    });
+  });
+  return headers;
+}
+
+export function serializeCsv(rows: Record<string, unknown>[]): string {
+  if (rows.length === 0) {
+    return '';
+  }
+  const headers = collectHeaders(rows);
+  const headerLine = headers.map(escapeCsvValue).join(',');
+  const dataLines = rows.map((row) =>
+    headers.map((header) => escapeCsvValue(row[header])).join(','),
+  );
+  return [headerLine, ...dataLines].join('\n');
+}
+
+export function loadDataset(datasetPath: string): { data: any[]; format: DatasetFormat } {
+  const datasetFullPath = path.resolve(datasetPath);
+  const rawData = fs.readFileSync(datasetFullPath, 'utf8');
+  const format = detectDatasetFormat(datasetFullPath);
+
+  if (format === 'json') {
+    const parsed = JSON.parse(rawData);
+    if (!Array.isArray(parsed)) {
+      throw new Error('Dataset JSON must be an array of records');
+    }
+    return { data: parsed, format };
+  }
+
+  const parsed = parseCsv(rawData);
+  return { data: parsed, format };
+}
+
+export function writeDataset(
+  outputDir: string,
+  rows: RunReturn[],
+  format: DatasetFormat,
+  config: Config,
+): void {
+  const outputDirPath = path.resolve(outputDir);
+  fs.mkdirSync(outputDirPath, { recursive: true });
+
+  const datasetFilename = DATASET_FILENAMES[format];
+  const datasetPath = path.join(outputDirPath, datasetFilename);
+  const configPath = path.join(outputDirPath, 'config.json');
+
+  if (format === 'json') {
+    fs.writeFileSync(datasetPath, JSON.stringify(rows, null, 4), 'utf8');
+  } else {
+    const csvContent = serializeCsv(rows);
+    fs.writeFileSync(datasetPath, csvContent, 'utf8');
+  }
+
+  fs.writeFileSync(configPath, JSON.stringify(config, null, 4), 'utf8');
+
+  console.info(`Output written to ${datasetPath}`);
+  console.info(`Config written to ${configPath}`);
+}
+
 class CLIHandler {
   private run: (...args: any[]) => Promise<any>;
 
@@ -46,10 +190,7 @@ class CLIHandler {
     const options = program.opts();
     const { datasetPath, outputDir } = options;
 
-    // Load dataset
-    const datasetFullPath = path.resolve(datasetPath);
-    const rawData = fs.readFileSync(datasetFullPath, 'utf8');
-    const dataset = JSON.parse(rawData);
+    const { data: dataset, format } = loadDataset(datasetPath);
 
     // Process each item in the dataset dynamically
     Promise.all(
@@ -64,7 +205,7 @@ class CLIHandler {
          * Wait for all rows to be run
          * Write results now to output dir or log to console
          */
-        this.writeOutput(results, outputDir);
+        this.writeOutput(results, outputDir, format);
         console.log('Results processing completed. Check console for output.');
       })
       .catch((err) => {
@@ -72,24 +213,13 @@ class CLIHandler {
       });
   }
 
-  private writeOutput(results: RunReturn[], outputDir: string) {
+  private writeOutput(results: RunReturn[], outputDir: string, format: DatasetFormat) {
     const config: Config = {
       metadata: { outputTimestamp: Date.now() },
       outputColumnName: 'output',
     };
 
-    // Construct an output directory {outputDir}/{datasetName}/
-    const outputDirPath = path.resolve(outputDir);
-    fs.mkdirSync(outputDirPath, { recursive: true });
-
-    const datasetPath = path.join(outputDirPath, 'dataset.json');
-    const configPath = path.join(outputDirPath, 'config.json');
-
-    fs.writeFileSync(datasetPath, JSON.stringify(results, null, 4), 'utf8');
-    fs.writeFileSync(configPath, JSON.stringify(config, null, 4), 'utf8');
-
-    console.info(`Output written to ${datasetPath}`);
-    console.info(`Config written to ${configPath}`);
+    writeDataset(outputDir, results, format, config);
   }
 }
 
diff --git a/tests/cli-dataset.test.ts b/tests/cli-dataset.test.ts
new file mode 100644
index 00000000..8cb501d3
--- /dev/null
+++ b/tests/cli-dataset.test.ts
@@ -0,0 +1,109 @@
+import * as fs from 'fs';
+import * as os from 'os';
+import * as path from 'path';
+
+import {
+  Config,
+  loadDataset,
+  parseCsv,
+  serializeCsv,
+  writeDataset,
+} from '../src/lib/core/cli';
+
+describe('CLI dataset helpers', () => {
+  const makeTempDir = (): string => fs.mkdtempSync(path.join(os.tmpdir(), 'ol-cli-'));
+    
+  it('loads JSON datasets', () => {
+    const dir = makeTempDir();
+    const filePath = path.join(dir, 'dataset.json');
+    const payload = [{ a: 1 }, { a: 2 }];
+    fs.writeFileSync(filePath, JSON.stringify(payload), 'utf8');
+
+    const { data, format } = loadDataset(filePath);
+
+    expect(format).toBe('json');
+    expect(data).toEqual(payload);
+  });
+
+  it('loads CSV datasets', () => {
+    const dir = makeTempDir();
+    const filePath = path.join(dir, 'dataset.csv');
+    const payload = 'id,name\n1,Alice\n2,Bob\n';
+    fs.writeFileSync(filePath, payload, 'utf8');
+
+    const { data, format } = loadDataset(filePath);
+
+    expect(format).toBe('csv');
+    expect(data).toEqual([
+      { id: '1', name: 'Alice' },
+      { id: '2', name: 'Bob' },
+    ]);
+  });
+
+  it('writes JSON outputs and config', () => {
+    const dir = makeTempDir();
+    const config: Config = {
+      metadata: { outputTimestamp: 123 },
+      outputColumnName: 'output',
+    };
+    const rows = [
+      { input: 'a', output: 'x', otherFields: {}, latency: 10 },
+      { input: 'b', output: 'y', otherFields: {}, latency: 20 },
+    ];
+
+    writeDataset(dir, rows, 'json', config);
+
+    const datasetPath = path.join(dir, 'dataset.json');
+    const configPath = path.join(dir, 'config.json');
+    expect(fs.existsSync(datasetPath)).toBe(true);
+    expect(fs.existsSync(configPath)).toBe(true);
+
+    const writtenDataset = JSON.parse(fs.readFileSync(datasetPath, 'utf8'));
+    const writtenConfig = JSON.parse(fs.readFileSync(configPath, 'utf8'));
+
+    expect(writtenDataset).toEqual(rows);
+    expect(writtenConfig).toEqual(config);
+  });
+
+  it('writes CSV outputs preserving headers and quoting values', () => {
+    const dir = makeTempDir();
+    const config: Config = {
+      metadata: { outputTimestamp: 456 },
+      outputColumnName: 'output',
+    };
+    const rows = [
+      { prompt: 'hello', output: 'hi', latency: 1 },
+      { prompt: 'quote, "comma"', output: 'ok', latency: 2 },
+    ];
+
+    writeDataset(dir, rows, 'csv', config);
+
+    const datasetPath = path.join(dir, 'dataset.csv');
+    const configPath = path.join(dir, 'config.json');
+    expect(fs.existsSync(datasetPath)).toBe(true);
+    expect(fs.existsSync(configPath)).toBe(true);
+
+    const csvContent = fs.readFileSync(datasetPath, 'utf8');
+    const parsed = parseCsv(csvContent);
+
+    expect(parsed).toEqual([
+      { prompt: 'hello', output: 'hi', latency: '1' },
+      { prompt: 'quote, "comma"', output: 'ok', latency: '2' },
+    ]);
+  });
+
+  it('serializes and parses CSV round-trip', () => {
+    const rows = [
+      { a: '1', b: 'text' },
+      { a: '2', b: 'text, with comma' },
+    ];
+    const serialized = serializeCsv(rows);
+    const parsed = parseCsv(serialized);
+
+    expect(parsed).toEqual([
+      { a: '1', b: 'text' },
+      { a: '2', b: 'text, with comma' },
+    ]);
+  });
+});
+